The coalescer does not coalesce a virtual register to a physical register if any of the physical register's sub-register live intervals overlaps with the virtual register. This is overly conservative. It prevents a extract_subreg from being coalesced away:

v1024 = EDI  // not killed
      =
      = EDI

One possible solution is for the coalescer to examine the sub-register live intervals in the same manner as the physical register. Another possibility is to examine defs and uses (when needed) of sub-registers. Both solutions are too expensive. For now, look for "short virtual intervals" and scan instructions to look for conflict instead.

This is a small win on x86-64. e.g. It shaves 403.gcc by ~80 instructions.

llvm-svn: 61847

aggregate types. Don't increment the current index after reaching
the end of a struct, as it will already be pointing at
one-past-the end. This fixes PR3288.

llvm-svn: 61828

llvm-svn: 61822

llvm-svn: 61819

two address instructions.  We need to keep track of things we've processed AS USES
independetly of whether we've processed them as defs.

This fixes all known miscompilations when reconstruction is turned on.

llvm-svn: 61802

argument. This doesn't affect current functionality.

llvm-svn: 61779

AddPseudoTwoAddrDeps. This lets the scheduling infrastructure
avoid recalculating node heights. In very large testcases this
was a major bottleneck. Thanks to Roman Levenstein for finding
this!

As a side effect, fold-pcmpeqd-0.ll is now scheduled better
and it no longer requires spilling on x86-32.

llvm-svn: 61778

Construct subprogram DIEs using DebugInfo.

llvm-svn: 61772

Construct global variable DIEs using DebugInfo.

llvm-svn: 61771

Construct compile unit dies using DebugInfo.

llvm-svn: 61768

llvm-svn: 61765

with latency 0, since it doesn't affect the depth or height.

llvm-svn: 61762

Add methods to add source location info in a DIE.

llvm-svn: 61761

llvm-svn: 61757

llvm-svn: 61741

own OpActionsCapacity magic number; it can just use ISD::BUILTIN_OP_END,
as long as it takes care to round up when needed.

llvm-svn: 61733

llvm-svn: 61731

Construct stuct field DIEs.

llvm-svn: 61729

llvm-svn: 61726

llvm-svn: 61724

problem, rather than fixing it.  The problem has now been fixed the right way.

llvm-svn: 61723

llvm-svn: 61715

llvm-svn: 61714

llvm-svn: 61711

subsume ConstructBasicType().

llvm-svn: 61709

squash warnings.

llvm-svn: 61707

of the pointer to the global.

llvm-svn: 61630

llvm::PATypeHolder::get() method when LLVM is self-hosted in Release
mode. Before the parser changed, there was a definition of llvm::PAHolder::get()
in llvmAsmParser.y. This was probably a bug that no-one noticed.

Explicitly #include the Type.h file as a temporary fix for now.

llvm-svn: 61620

llvm-svn: 61613

llvm-svn: 61612

instructions to avoid copies, because TwoAddressInstructionPass
also does this optimization.  The scheduler's version didn't
account for live-out values, which resulted in spurious commutes
and missed opportunities.

Now, TwoAddressInstructionPass handles all the opportunities,
instead of just those that the scheduler missed. The result is
usually the same, though there are occasional trivial differences
resulting from the avoidance of spurious commutes.

llvm-svn: 61611

and BRCOND conditions.  Reorder a few methods while
there.

llvm-svn: 61547

llvm-svn: 61545

promote from i1 all the way up to the canonical SetCC type.
In order to discover an appropriate type to use, pass
MVT::Other to getSetCCResultType.  In order to be able to
do this, change getSetCCResultType to take a type as an
argument, not a value (this is also more logical).

llvm-svn: 61542

llvm-svn: 61514

Darwin doesn't. Make this optional for platforms.

llvm-svn: 61484

(quad) on a 64-bit platform. This fixes a problem with EH frames on Darwin.

llvm-svn: 61483

llvm-svn: 61465

llvm-svn: 61463

llvm-svn: 61460